This paper: (1) examines the problems of teacher evaluation by college students; (2) underlines the difficulties of indiscriminant acceptance of raw data without considering the factors of teaching style, learning modalities, and educational goals; and (3) urges that evaluative instruments be used primarily for diagnostic purposes, for the improvement of learning, of teaching, and of the learner-teacher relationship. Three problem areas in particular are discussed: (1) the dilemma of teaching style, involving questions of teacher vs. student motivated learning; (2) the dilemma of recommendations to individual teachers, stressing the need for constructive evaluation, conditional upon the instructor's views of what is appropriate, practical, and desirable to him in respect to his instructional goals, and (3) the dilemma of designing valid evaluation instruments, instruments that take into account factors such as teaching modes, teacher expectations, learner performance, student workload both in and outside the class, and instrument internal consistency. Discussion of five factor analysis clusters determined by Hildebrand as characteristic of best/worst teachers illustrates further difficulties in evaluation procedures; the author offers remedial suggestions. (MB)